Paper Info
Title
An Improved N-Step Value Gradient Learning Adaptive Dynamic Programming Algorithm for Online Learning.
Abstract
In problems with complex dynamics and challenging state spaces, the dual heuristic programming (DHP) algorithm has been shown theoretically and experimentally to perform well. This was recently extended by an approach called value gradient learning (VGL). VGL was inspired by a version of temporal difference (TD) learning that uses eligibility traces. The eligibility traces create an exponential decay of older observations with a decay parameter ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\lambda $ </tex-math></inline-formula> ). This approach is known as TD( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\lambda $ </tex-math></inline-formula> ), and its DHP extension is known as VGL( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\lambda $ </tex-math></inline-formula> ), where VGL(0) is identical to DHP. VGL has presented convergence and other desirable properties, but it is primarily useful for batch learning. Online learning requires an eligibility-trace-work-space matrix, which is not required for the batch learning version of VGL. Since online learning is desirable for many applications, it is important to remove this computational and memory impediment. This paper introduces a dual-critic version of VGL, called <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$N$ </tex-math></inline-formula> -step VGL (NSVGL), that does not need the eligibility-trace-work-space matrix, thereby allowing online learning. Furthermore, this combination of critic networks allows an NSVGL algorithm to learn faster. The first critic is similar to DHP, which is adapted based on TD(0) learning, while the second critic is adapted based on a gradient of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$n$ </tex-math></inline-formula> -step TD( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\lambda $ </tex-math></inline-formula> ) learning. Both networks are combined to train an actor network. The combination of feedback signals from both critic networks provides an optimal decision faster than traditional adaptive dynamic programming (ADP) via mixing current information and event history. Convergence proofs are provided. Gradients of one- and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$n$ </tex-math></inline-formula> -step value functions are monotonically nondecreasing and converge to the optimum. Two simulation case studies are presented for NSVGL to show their superior performance.
Year
DOI
Venue
2020
10.1109/TNNLS.2019.2919338
IEEE transactions on neural networks and learning systems
Keywords
DocType
Volume
Adaptive dynamic programming (ADP),convergence analysis,eligibility traces,online learning,reinforcement learning,temporal difference (TD),value gradient learning (VGL)
Journal
31
Issue
ISSN
Citations 
4
2162-237X
1
PageRank 
References 
Authors
0.35
27
2
Name
Order
Citations
PageRank
Seaar Al-Dabooni1101.82
Wunsch II Donald C.2135491.73